so i have recently beaten the WR in a game and wanted to try and imrpove it but it's been so hard for me to improve the WR that i wanna see what my chances are of beating my time before i take another possible 5+ months to try and improve my time XD

is there any sort of way to calculate how many runs it might take me using attempts made and runs finished?

Realistically I think you could only accurately calculate such a thing if the run is mostly RNG based (like, if something with a 1% chance of happening needs to happen for you to get WR, then there's some probability math you can use to get a rough idea of how many attempts you might need). But if it's mostly skill-based, your skill as a player and how fast you're improving are extremely difficult to quantify in a useful way for this.

Instead of trying to math out an equation for this and then brute-forcing attempts, you might need to just change gears in terms of what you're practicing and how often you're practicing. Or perhaps it's just time to take a break from the category and try something else for a bit, counterintuitively taking a break can actually help you improve. I'm afraid I can't give any more specific advice though since I don't know much about the game in question.

I think it might be possible to find a fitting distribution to any given split if there's not too many different types of tricks/RNG involved. Like in the RNG example: if everything that is in that split can be approximated by normal distribution, then the whole split can be treated so. With the platforming example it might be very individual: let's say you have the history all your attempts of a split. Then maybe by looking at a plot of that data you might see that for a particular trick you get it 90% of the times right the 1st time then have a gap before maybe 9% of the times to get it 2nd try and so on.

But I'd guess you'd have to be very careful with what you combine in a split. If you have normal distributed & tricks with maybe weird distributions in a split then the resulting distribution might be very weird. And then calculating things with such a "bad" data quality is probably not giving any desired result.